add help information

eval_w.py

@@ -21,11 +21,11 @@
 import functools
 
 if __name__ == "__main__":
-    parser = argparse.ArgumentParser(description='Evaluating CHAR-CW-BLSTM-CRF')
-    parser.add_argument('--load_arg', default='./checkpoint/soa/check_wc_p_char_lstm_crf.json')
-    parser.add_argument('--load_check_point', default='./checkpoint/soa/check_wc_p_char_lstm_crf.model')
-    parser.add_argument('--gpu',type=int, default=0)
-    parser.add_argument('--eva_matrix', choices=['a', 'fa'], default='fa')
+    parser = argparse.ArgumentParser(description='Evaluating BLSTM-CRF')
+    parser.add_argument('--load_arg', default='./checkpoint/soa/check_wc_p_char_lstm_crf.json', help='arg json file path')
+    parser.add_argument('--load_check_point', default='./checkpoint/soa/check_wc_p_char_lstm_crf.model', help='checkpoint path')
+    parser.add_argument('--gpu',type=int, default=0, help='gpu id')
+    parser.add_argument('--eva_matrix', choices=['a', 'fa'], default='fa', help='use f1 and accuracy or accuracy alone')
     args = parser.parse_args()
 
     with open(args.load_arg, 'r') as f:

eval_wc.py

@@ -21,11 +21,11 @@
 import functools
 
 if __name__ == "__main__":
-    parser = argparse.ArgumentParser(description='Evaluating CHAR-CW-BLSTM-CRF')
-    parser.add_argument('--load_arg', default='./checkpoint/soa/check_wc_p_char_lstm_crf.json')
-    parser.add_argument('--load_check_point', default='./checkpoint/soa/check_wc_p_char_lstm_crf.model')
-    parser.add_argument('--gpu',type=int, default=0)
-    parser.add_argument('--eva_matrix', choices=['a', 'fa'], default='fa')
+    parser = argparse.ArgumentParser(description='Evaluating LM-BLSTM-CRF')
+    parser.add_argument('--load_arg', default='./checkpoint/soa/check_wc_p_char_lstm_crf.json', help='path to arg json')
+    parser.add_argument('--load_check_point', default='./checkpoint/soa/check_wc_p_char_lstm_crf.model',, help='path to model checkpoint file')
+    parser.add_argument('--gpu',type=int, default=0, help='gpu id')
+    parser.add_argument('--eva_matrix', choices=['a', 'fa'], default='fa', help='use f1 and accuracy or f1 alone')
     args = parser.parse_args()
 
     with open(args.load_arg, 'r') as f:

model/__init__.py

@@ -1,7 +1,6 @@
 __author__ = "Liyuan Liu and Frank Xu"
 __credits__ = ["Liyuan Liu", "Frank Xu", "Jingbo Shang"]
 
-__license__ = "GPL"
-__version__ = "1.0.1"
+__license__ = "Apache License 2.0"
 __maintainer__ = "Liyuan Liu"
 __email__ = "llychinalz@gmail.com"

model/crf.py

@@ -247,7 +247,9 @@ def __init__(self, tagset_size, start_tag, end_tag, average_batch=True):
         self.crit = nn.CrossEntropyLoss(size_average=self.average_batch)
 
     def forward(self, scores, target, current):
-        """greedy decode loss, i.e., although its for CRF Layer, we calculate the loss as :math:`\prod_{j=1}^n p(l_{j+1}|z_{j+1}, l_{j})` instead of :math:`\frac{ \prod_{j=1}^n \phi(\hat{y}_{j-1}, \hat{y}_j, \z_j) } { \sum_{\y' \in \Y(\Z)} \prod_{j=1}^n \phi(y'_{j-1}, y'_j, \z_j)}`
+        """greedy decode loss, i.e., although its for CRF Layer, we calculate the loss as :math:`\prod_{j=1}^n p(l_{j+1}|z_{j+1}, l_{j})` 
+        instead of 
+        :math:`\frac{ \prod_{j=1}^n \phi(\hat{y}_{j-1}, \hat{y}_j, \mathbf{z}_j) } { \sum_{\mathbf{y}' \in \mathbf{Y}(\mathbf{Z})} \prod_{j=1}^n \phi(y'_{j-1}, y'_j, \mathbf{z}_j)}`
 
         args: 
             scores: crf scores, of size Word_Seq_len * Batch_size * target_size_from * target_size_to
@@ -284,7 +286,8 @@ def __init__(self, tagset_size, start_tag, end_tag, average_batch=True):
 
     def forward(self, scores, target, mask):
         """
-        viterbi decode loss, \frac{ \prod_{j=1}^n \phi(\hat{y}_{j-1}, \hat{y}_j, \z_j) } { \sum_{\y' \in \Y(\Z)} \prod_{j=1}^n \phi(y'_{j-1}, y'_j, \z_j)}
+        viterbi decode loss, 
+        :math:`\sum_{j=1}^n \log(\phi(\hat{y}_{j-1}, \hat{y}_j, \mathbf{z}_j)) - \log(\sum_{\mathbf{y}' \in \mathbf{Y}(\mathbf{Z})} \prod_{j=1}^n \phi(y'_{j-1}, y'_j, \mathbf{z}_j) )`
 
         args:
             scores: crf scores, of size seq_len * bat_size * target_size_from * target_size_to
@@ -314,9 +317,7 @@ def forward(self, scores, target, mask):
             # previous to_target is current from_target
             # partition: previous results log(exp(from_target)), #(batch_size * from_target)
             # cur_values: bat_size * from_target * to_target
-            cur_values = cur_values + partition.contiguous().view(bat_size, self.tagset_size, 1).expand(bat_size,
-                                                                                                        self.tagset_size,
-                                                                                                        self.tagset_size)
+            cur_values = cur_values + partition.contiguous().view(bat_size, self.tagset_size, 1).expand(bat_size, self.tagset_size, self.tagset_size)
             cur_partition = utils.log_sum_exp(cur_values, self.tagset_size)
                   # (bat_size * from_target * to_target) -> (bat_size * to_target)
             partition = utils.switch(partition, cur_partition,

model/evaluator.py

@@ -1,3 +1,11 @@
+"""
+.. module:: evaluator
+    :synopsis: evaluator for f1 and acc
+ 
+.. moduleauthor:: Liyuan Liu, Frank Xu
+"""
+
+
 import torch
 import numpy as np
 import itertools

train_w.py

@@ -21,35 +21,35 @@
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description='Learning with BLSTM-CRF')
-    parser.add_argument('--rand_embedding', action='store_true')
-    parser.add_argument('--emb_file', default='./embedding/glove.6B.100d.txt')
-    parser.add_argument('--train_file', default='./data/ner2003/eng.train.iobes')
-    parser.add_argument('--dev_file', default='./data/ner2003/eng.testa.iobes')
-    parser.add_argument('--test_file', default='./data/ner2003/eng.testb.iobes')
-    parser.add_argument('--gpu', type=int, default=0)
-    parser.add_argument('--batch_size', type=int, default=10)
-    parser.add_argument('--unk', default='unk')
-    parser.add_argument('--checkpoint', default='./checkpoint/')
-    parser.add_argument('--hidden', type=int, default=100)
-    parser.add_argument('--drop_out', type=float, default=0.5)
-    parser.add_argument('--epoch', type=int, default=200)
-    parser.add_argument('--start_epoch', type=int, default=0)
-    parser.add_argument('--caseless', action='store_true')
-    parser.add_argument('--embedding_dim', type=int, default=100)
-    parser.add_argument('--layers', type=int, default=1)
-    parser.add_argument('--lr', type=float, default=0.01)
-    parser.add_argument('--lr_decay', type=float, default=0.001)
-    parser.add_argument('--fine_tune', action='store_false')
-    parser.add_argument('--load_check_point', default='')
-    parser.add_argument('--load_opt', action='store_true')
-    parser.add_argument('--update', choices=['sgd', 'adam'], default='sgd')
-    parser.add_argument('--momentum', type=float, default=0.9)
-    parser.add_argument('--clip_grad', type=float, default=5.0)
-    parser.add_argument('--small_crf', action='store_false')
-    parser.add_argument('--mini_count', type=float, default=5)
-    parser.add_argument('--eva_matrix', choices=['a', 'fa'], default='fa')
-    parser.add_argument('--patience', type=int, default=15)
-    parser.add_argument('--least_iters', type=int, default=50)
+    parser.add_argument('--rand_embedding', action='store_true', help='random initialize word embedding')
+    parser.add_argument('--emb_file', default='./embedding/glove.6B.100d.txt', help='path to pre-trained embedding')
+    parser.add_argument('--train_file', default='./data/ner2003/eng.train.iobes', help='path to training file')
+    parser.add_argument('--dev_file', default='./data/ner2003/eng.testa.iobes', help='path to development file')
+    parser.add_argument('--test_file', default='./data/ner2003/eng.testb.iobes', help='path to test file')
+    parser.add_argument('--gpu', type=int, default=0, help='gpu id, set to -1 if use cpu mode')
+    parser.add_argument('--batch_size', type=int, default=10, help='batch size (10)')
+    parser.add_argument('--unk', default='unk', help='unknow-token in pre-trained embedding')
+    parser.add_argument('--checkpoint', default='./checkpoint/', help='path to checkpoint prefix')
+    parser.add_argument('--hidden', type=int, default=100, help='hidden dimension')
+    parser.add_argument('--drop_out', type=float, default=0.5, help='dropout ratio')
+    parser.add_argument('--epoch', type=int, default=200, help='maximum epoch number')
+    parser.add_argument('--start_epoch', type=int, default=0, help='start epoch idx')
+    parser.add_argument('--caseless', action='store_true', help='caseless or not')
+    parser.add_argument('--embedding_dim', type=int, default=100, help='dimension for word embedding')
+    parser.add_argument('--layers', type=int, default=1, help='number of lstm layers')
+    parser.add_argument('--lr', type=float, default=0.01, help='initial learning rate')
+    parser.add_argument('--lr_decay', type=float, default=0.001, help='decay ratio of learning rate')
+    parser.add_argument('--fine_tune', action='store_false', help='fine tune pre-trained embedding dictionary')
+    parser.add_argument('--load_check_point', default='', help='path of checkpoint')
+    parser.add_argument('--load_opt', action='store_true', help='load optimizer from ')
+    parser.add_argument('--update', choices=['sgd', 'adam'], default='sgd', help='optimizer method')
+    parser.add_argument('--momentum', type=float, default=0.9, help='momentum for sgd')
+    parser.add_argument('--clip_grad', type=float, default=5.0, help='grad clip at')
+    parser.add_argument('--small_crf', action='store_false', help='use small crf instead of large crf, refer model.crf module for more details')
+    parser.add_argument('--mini_count', type=float, default=5, help='thresholds to replace rare words with <unk>')
+    parser.add_argument('--eva_matrix', choices=['a', 'fa'], default='fa', help='use f1 and accuracy or accuracy alone')
+    parser.add_argument('--patience', type=int, default=15, help='patience for early stop')
+    parser.add_argument('--least_iters', type=int, default=50, help='at least train how many epochs before stop')
     args = parser.parse_args()
 
     if args.gpu >= 0:

train_wc.py

@@ -24,42 +24,42 @@ def eprint(*args, **kwargs):
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description='Learning with LM-LSTM-CRF together with Language Model')
-    parser.add_argument('--rand_embedding', action='store_true')
-    parser.add_argument('--emb_file', default='./embedding/glove.6B.100d.txt')
-    parser.add_argument('--train_file', default='./data/ner2003/eng.train.iobes')
-    parser.add_argument('--dev_file', default='./data/ner2003/eng.testa.iobes')
-    parser.add_argument('--test_file', default='./data/ner2003/eng.testb.iobes')
-    parser.add_argument('--gpu', type=int, default=0)
-    parser.add_argument('--batch_size', type=int, default=10)
-    parser.add_argument('--unk', default='unk')
-    parser.add_argument('--char_hidden', type=int, default=300)
-    parser.add_argument('--word_hidden', type=int, default=300)
-    parser.add_argument('--drop_out', type=float, default=0.5)
-    parser.add_argument('--epoch', type=int, default=200)
-    parser.add_argument('--start_epoch', type=int, default=0)
-    parser.add_argument('--checkpoint', default='./checkpoint/')
-    parser.add_argument('--caseless', action='store_true')
-    parser.add_argument('--char_dim', type=int, default=30)
-    parser.add_argument('--word_dim', type=int, default=100)
-    parser.add_argument('--char_layers', type=int, default=1)
-    parser.add_argument('--word_layers', type=int, default=1)
-    parser.add_argument('--lr', type=float, default=0.01)
-    parser.add_argument('--lr_decay', type=float, default=0.05)
-    parser.add_argument('--fine_tune', action='store_false')
-    parser.add_argument('--load_check_point', default='')
-    parser.add_argument('--load_opt', action='store_true')
-    parser.add_argument('--update', choices=['sgd', 'adam'], default='sgd')
-    parser.add_argument('--momentum', type=float, default=0.9)
-    parser.add_argument('--clip_grad', type=float, default=5.0)
-    parser.add_argument('--small_crf', action='store_false')
-    parser.add_argument('--mini_count', type=float, default=5)
-    parser.add_argument('--lambda0', type=float, default=1)
-    parser.add_argument('--co_train', action='store_true')
-    parser.add_argument('--patience', type=int, default=15)
-    parser.add_argument('--high_way', action='store_true')
-    parser.add_argument('--highway_layers', type=int, default=1)
-    parser.add_argument('--eva_matrix', choices=['a', 'fa'], default='fa')
-    parser.add_argument('--least_iters', type=int, default=50)
+    parser.add_argument('--rand_embedding', action='store_true', help='random initialize word embedding')
+    parser.add_argument('--emb_file', default='./embedding/glove.6B.100d.txt', help='path to pre-trained embedding')
+    parser.add_argument('--train_file', default='./data/ner2003/eng.train.iobes', help='path to training file')
+    parser.add_argument('--dev_file', default='./data/ner2003/eng.testa.iobes', help='path to development file')
+    parser.add_argument('--test_file', default='./data/ner2003/eng.testb.iobes', help='path to test file')
+    parser.add_argument('--gpu', type=int, default=0, help='gpu id')
+    parser.add_argument('--batch_size', type=int, default=10, help='batch_size')
+    parser.add_argument('--unk', default='unk', help='unknow-token in pre-trained embedding')
+    parser.add_argument('--char_hidden', type=int, default=300, help='dimension of char-level layers')
+    parser.add_argument('--word_hidden', type=int, default=300, help='dimension of word-level layers')
+    parser.add_argument('--drop_out', type=float, default=0.5, help='dropout ratio')
+    parser.add_argument('--epoch', type=int, default=200, help='maximum epoch number')
+    parser.add_argument('--start_epoch', type=int, default=0, help='start point of epoch')
+    parser.add_argument('--checkpoint', default='./checkpoint/', help='checkpoint path')
+    parser.add_argument('--caseless', action='store_true', help='caseless or not')
+    parser.add_argument('--char_dim', type=int, default=30, help='dimension of char embedding')
+    parser.add_argument('--word_dim', type=int, default=100, help='dimension of word embedding')
+    parser.add_argument('--char_layers', type=int, default=1, help='number of char level layers')
+    parser.add_argument('--word_layers', type=int, default=1, help='number of word level layers')
+    parser.add_argument('--lr', type=float, default=0.01, help='initial learning rate')
+    parser.add_argument('--lr_decay', type=float, default=0.05, help='decay ratio of learning rate')
+    parser.add_argument('--fine_tune', action='store_false', help='fine tune the diction of word embedding or not')
+    parser.add_argument('--load_check_point', default='', help='path previous checkpoint that want to be loaded')
+    parser.add_argument('--load_opt', action='store_true', help='also load optimizer from the checkpoint')
+    parser.add_argument('--update', choices=['sgd', 'adam'], default='sgd', help='optimizer choice')
+    parser.add_argument('--momentum', type=float, default=0.9, help='momentum for sgd')
+    parser.add_argument('--clip_grad', type=float, default=5.0, help='clip grad at')
+    parser.add_argument('--small_crf', action='store_false', help='use small crf instead of large crf, refer model.crf module for more details')
+    parser.add_argument('--mini_count', type=float, default=5, help='thresholds to replace rare words with <unk>')
+    parser.add_argument('--lambda0', type=float, default=1, help='lambda0')
+    parser.add_argument('--co_train', action='store_true', help='cotrain language model')
+    parser.add_argument('--patience', type=int, default=15, help='patience for early stop')
+    parser.add_argument('--high_way', action='store_true', help='use highway layers')
+    parser.add_argument('--highway_layers', type=int, default=1, help='number of highway layers')
+    parser.add_argument('--eva_matrix', choices=['a', 'fa'], default='fa', help='use f1 and accuracy or accuracy alone')
+    parser.add_argument('--least_iters', type=int, default=50, help='at least train how many epochs before stop')
     args = parser.parse_args()
 
     if args.gpu >= 0: